Recently, chromosomal rearrangements involving receptor tyrosine kinases (RTKs) have been described in common epithelial malignancies, including nonsmall cell lung cancer (NSCLC), colorectal cancer, and breast cancer. One of these RTKs, c-ros oncogene 1, receptor tyrosine kinase (ROS1), has been identified as a driver mutation in NSCLC, because its inhibition by crizotinib, an anaplastic lymphoma receptor tyrosine kinase (ALK)/met proto-oncogene hepatocyte growth factor receptor (MET)/ROS1 inhibitor, led to significant tumor shrinkage in ROS1-rearranged NSCLC. Currently, only human epidermal growth factor 2 (HER2)-targeted therapy in combination with chemotherapy has been successful in significantly prolonging the survival of patients with advanced gastric cancer (GC). There is a need for the discovery of additional novel targets in GC.
Anti-ROS1 immunohistochemistry (IHC) was used to screen 495 GC samples and was followed by simultaneous ROS1 break-apart fluorescence in situ hybridization (FISH) and reverse transcriptase-polymerase chain reaction (RT-PCR) analyses in IHC-positive samples. Fusion partners in ROS1-rearranged GC were determined by RT-PCR. In all 495 samples, HER2 amplification was identified with FISH, and MET expression was identified by IHC.
Twenty-three tumor samples were ROS1 IHC-positive. Three of 23 patients were ROS1 FISH positive, HER2 FISH negative, and negative for MET overexpression; and 2 of those 3 patients harbored a solute carrier family 34 (sodium phosphate), member 2 (SLC34A2)-ROS1 fusion transcripts. No fusion partner was identified in the third patient. Both patients who had SLC34A2-ROS1 transcripts had poorly differentiated histology with recurrence and death within 2 years of curative surgery. ROS1 IHC-positive status was not identified as an independent prognostic factor for overall survival.
In 2008, 989,600 cases of gastric cancer (GC) were projected, and GC was responsible for 738,000 deaths globally, making GC the second most common cause of cancer death worldwide.1 Among developing countries, GC is the third most common cause of cancer death.1 Even among developed countries, death from GC ranks as the fourth most common cause of cancer death after lung, colorectal, and breast cancers, in that order.1 Although many potentially “actionable” receptor tyrosine kinases (RTKs) have been identified in GC, such as amplification of the mesenchymal epithelial transition (MET) receptor, the epidermal growth factor receptor (EGFR), human epidermal growth factor receptor 2 (HER2), and fibroblastic growth factor receptor (FGFR),2 to date, only trastuzumab, an antibody against HER2, in combination with chemotherapy has been successful in prolonging the survival of patients with GC.3 ROS1 is another human RTK receptor that has been found to be rearranged in glioblastoma multiforme,4 nonsmall cell lung cancer (NSCLC),5-10 and cholangiocarcinoma.11 There have been encouraging preliminary reports of the clinical activity of crizotinib, an oral MET/anaplastic lymphoma kinase (ALK) inhibitor, in ROS1-rearranged NSCLC,12 indicating that ROS1 rearrangement is a driver mutation in NSCLC; thus, there is significant interest in ROS1-rearranged tumors. In this report, our objective was to investigate whether ROS1 is rearranged in GC. Because ROS1 is generally not expressed in stomach,8, 13 we reasoned that any expression of ROS1 protein in GC is likely aberrant and may represent gene amplification or translocation. We used an approach similar to that of Sugawara et al., who identified 2 cases of ALK-rearranged nonclear cell renal cell carcinoma using an initial immunohistochemistry (IHC) screening approach followed by confirmation using break-apart fluorescence in situ hybridization (FISH) and reverse transcriptase-polymerase chain reaction (RT-PCR) analyses.14
MATERIALS AND METHODS
Patients and Tissue Microarrays
From 2000 and 2004, 5133 patients with gastric adenocarcinoma underwent surgical treatment with curative intent at Samsung Medical Center, Seoul, Korea. Among these patients, 1048 patients presented with locally advanced GC; and, of those 1048 patients, 495 met the following inclusion criteria: histologically confirmed adenocarcinoma of the stomach, availability of paraffin-embedded tumor blocks, surgical resection of tumor without residual disease, age ≥18 years, pathologic stage IB through IV disease without metastasis (M0) according to the sixth edition of the American Joint Committee on Cancer (AJCC) staging system, and complete surgical and treatment records. All patients had undergone curative surgical resection with extensive (D2) lymph node dissection and had available clinicopathologic characteristics, surgical records, treatment records, recurrence status, and vital status at the last follow-up date. The last date of follow-up was July 1, 2012. This retrospective study was undertaken after we received approval by the Samsung Medical Center Institutional Review Board.
The tumor samples were fixed in 10% buffered formalin, processed, and embedded in paraffin using the standard protocol. All hematoxylin and eosin-stained slides were reviewed, and representative areas were carefully selected and marked on individual paraffin blocks. Four 0.6mm tissue cores were taken from the representative region of each tumor specimen using Accumax (ISU Abxis, Seoul, Korea). Each tissue microarray contained 45 carcinoma samples and 3 nontumor controls.
After deparaffinization and rehydration, 4-μm sections on silane-coated slides were used for IHC. All procedures were performed using the Leica BOND-MAX System after antigen retrieval with Bond Epitope Retrieval Solution (Leica Microsystems, Wetzlar, Germany). Anti-ROS primary antibody corresponding to amino acids 39 through 57 of human ROS (Abcam, Cambridge, Mass; 1:80 dilution) was used, and labeling was performed using the Bond Max Automated Immunohistochemistry Vision Biosystem (Leica Microsystems). Subsequently, tissues were incubated with polymer for 10 minutes and developed with diaminobenzidine as a chromogen for 10 minutes. For interpretation of ROS1 expression, 3+ perinuclear staining was considered positive.8 Two independent pathologists (S.E.L. and K.M.K.) evaluated the results.
Analysis of ROS1 Rearrangement by Fluorescence in Situ Hybridization
FISH studies were performed on interphase nuclei that were present on formalin-fixed, paraffin-embedded tissue sections from ROS1 IHC-positive samples. The HCC78 cell line (which harbors a solute carrier family 34 [sodium phosphate], member 2 [SLC34A2]-ROS1 transcript) was used as positive control. Unstained, 1-μm sections were placed on electrostatically charged slides (SuperFrost; Fisher, Hampton, NH) and evaluated with an ROS1 (6q22) dual-color, break-apart probe (catalog no. FS0015; Abnova, Walnut, Calif) according to the manufacturer's instructions. The 5′ probe was approximately 720 kb long, the 3′ probe was approximately 250 kb long, and there was a 60-kb overlap between the 5′ and 3′ probes within the ROS1 gene. The hybridized slides were reviewed on an Olympus IX-50 microscope (Olympus, Tokyo, Japan) at ×100 magnification with oil immersion, using a 4′,6-diamindino-2-pheynylindole/Green/Red triple-band-pass filter set. The interpretation of intact and split signals of ROS1 was based on criteria for ROS1 positivity in NSCLC: at least 50 nonoverlapping cells examined, a separation between the 5′ and 3′ signals greater than the width of a 2-signal diameter, and tumors that harbored >15% cells with split signals or isolated 3′ signals.6
RNA Extraction and Reverse-Transcriptase Polymerase Chain Reaction
Total RNA was isolated from 23 ROS1 IHC-positive, formalin-fixed, paraffin-embedded tumor samples using the RNeasy Mini kit (Qiagen, Valencia, Calif). RT-PCR was conducted using a high-capacity cDNA Reverse Transcription Kit (catalog no. 4368814; Applied Biosystems, Foster City, Calif) according to the manufacturer's instructions. To reveal suspected ROS1 partner genes, a 5′ rapid amplification of cDNA ends (RACE) method using the SMARTRACE cDNA Amplification Kit (Clontech, Mountain View, Calif) was used according to the manufacturer's instructions, as previously described, with ROS1-5992R (5′-ATTTGCTCATCAGATGTGCCTC CTTCAG-3′) gene-specific reverse primers. ROS1 breakpoints were confirmed by using RT-PCR with primers for the tropomyosin 3 (TPM3), SLC34A2, syndecan 4 (SDC4), CD74 (major histocompatibility complex, class II invariant chain), and ezrin (EZR) genes combined with direct Sanger sequencing of PCR products. In addition, we also used a panel of PCR primers, including forward primers for SLC34A2 exon 4 and CD74 exon 6 each paired with reverse primers for ROS1 exon 32 and exon 34, as previously described. For positive controls, we used RNA from HCC78 cells, and we used RNA from ROS1-negative, formalin-fixed, paraffin-embedded gastric tumor samples for negative controls.
HER2 Fluorescence in Situ Hybridization and Immunohistochemistry for MET
FISH was performed using the dual-color, DNA-specific PathVysion HER-2 DNA Probe Kit (Abbott/Vysis, Downer's Grove, Ill), as described previously,15 and the HER2 gene was considered amplified when the FISH signal ratio of HER2 to chromosome 17 centromere (CEP17) was ≥2.0.16 For MET IHC, the Ventana CONFIRM antitotal-MET (SP44) rabbit monoclonal primary antibody (Ventana Medical Systems, Tucson, Ariz) was used with the Ventana Bench Mark XT automated slide-processing system. Two independent pathologists (S.Y.H. and K.M.K.) with no prior knowledge of clinicopathologic or molecular status evaluated the results. IHC results for MET were interpreted by using a scoring system that was applied in a previous phase 2 lung cancer trial (the 2 OAM4458g trial).17
For comparisons of clinicopathologic characteristics between patients with positive versus negative ROS1 IHC results, the Pearson chi-square test was used. Overall survival was calculated from the time of diagnosis to the time of death or last follow-up using the Kaplan-Meier method. Univariate and multivariate analyses were performed using Cox regression models. All analyses were conducted using the SPSS software package (version 20; SPSS, Inc., Chicago, Ill).
Identification of Positive ROS1 Immunohistochemistry in Gastric Cancer Samples
We performed IHC in 495 gastric adenocarcinoma patient samples and identified 23 samples (4%) that had strong positive staining for ROS1. The ROS1 staining pattern differed from the HER2 staining pattern in gastric adenocarcinoma (basolateral membranous staining). Instead, ROS1 was stained in the perinuclear area with dot-like accentuation, which is the pattern of staining observed in ROS1-rearranged NSCLC (Fig. 1). The clinicopathologic characteristics of the 23 patients with ROS1 IHC-positive GC are listed in Table 1. Patients who had ROS1 IHC-positive gastric adenocarcinoma were likely to have lower lymph node status (based on the AJCC sixth edition; P < .001) and were less likely to have lymphatic invasion (P = .021). Those who had ROS1 IHC-positive GC also tended to present with differentiated tumors (P = .017) and with the intestinal Lauren type of GC (P = .035). There was no difference in age at diagnosis, sex distribution, location of the tumor in the stomach, or AJCC stage at diagnosis between patients who had positive versus negative ROS1 IHC results (Table 2).
Table 1. Clinicopathologic Characteristics of 23 Patients With Positive ROS1 Immunohistochemistry Results
Identification of ROS1-Rearranged Gastric Adenocarcinoma by Break-Apart Fluorescence in Situ Hybridization and Reverse Transcriptase-Polymerase Chain Reaction
Next, we attempted to confirm the existence of ROS1 rearrangements in FISH analysis of 23 ROS1 IHC-positive GC specimens. All 23 specimens were tested with ROS1 FISH break-apart probes, and 3 samples met the criteria to be considered FISH-positive. The percentages of positive cells in the 3 samples that were positive for the break-apart ROS1 FISH probe were 18%, 22%, and 28%. There was no evidence of ROS1 polysomy in the 23 IHC-positive samples. All 3 samples revealed separation of the 5′ (green) and 3′ (red) signals, and no sample had isolated 3′ (red) signals detected.
Reverse Transcriptase-Polymerase Chain Reaction
RT-PCR was successful in identifying the fusion partner to ROS1 in 2 samples. In both samples, RT-PCR revealed that exon 4 of SLC34A2 was fused to exon 32 of ROS1 (Fig. 2). Primers that encoded for CD74, stearoyl-coenzyme A desaturase 4 (SCD4), TPM3, leucine-rich repeats and immunoglobulin-like domains 3 (LRIG3), and EZR were not successful in identifying any ROS1 fusions in the remaining 21 samples3.
HER2 Fluorescence in Situ Hybridization and MET Immunohistochemistry
HER2 amplification by FISH was detected in 62 of 495 GC samples (12.5%). MET overexpression, as defined by an IHC score of 3+, was detected in 9 of 495 GC samples (1.9%).
Case Presentations and Histologic Examinations of the 3 ROS1-Rearranged Gastric Cancers
The first patient was a Korean woman aged 60 years who presented initially with stage II (T2bN0M0; AJCC sixth edition), poorly differentiated (diffuse type), mucinous adenocarcinoma in 2001. The greatest dimension of the resected tumor was 6.5 cm. The tumor was located in the mid-body. There was no lymphovascular or perineural invasion identified in the surgically resected tumor. The patient underwent resection and received 5 months of 5-fluororacil-based adjuvant chemotherapy but developed a recurrence 15 months after completing chemoradiation therapy. The patient presented with malignant ascites that had peritoneal seeding at the time of recurrence and died of the disease 2 months later. The tumor was negative for HER2 gene amplification by FISH and negative for MET overexpression by IHC. Twenty-eight percent of tumor cells were positive in the ROS1 break-apart FISH analysis. The fusion partner in this tumor was SLC34A2.
The patient was a Korean woman aged 63 years who presented initially with stage IIIB (T2N3M0), poorly differentiated adenocarcinoma. The greatest dimension measured from the surgically resected tumor was 9 cm. The tumor was located in the antrum. Both lymphovascular and perineural invasion was identified in the tumor specimen. Similar to the first patient, this woman had a Lauren diffuse type tumor. The patient recurred with peritoneal seeding shortly after surgery; she declined palliative chemotherapy and died of disease 18 months later. The tumor was negative for HER2 amplification by FISH and had an IHC score of 1+ for MET expression. Eighteen percent of tumor cells were positive in the ROS1 break-apart FISH analysis. The fusion partner in this tumor was SLC34A2.
This patient was a Korean woman aged 62 years who underwent surgical resection. The percentage of tumor cells identified as positive in the ROS1 break-apart FISH analysis was 22%, and there was no fusion partner identified in this tumor. The patient remained alive without recurrence at the time of analysis.
There was no difference in overall survival between patients with positive versus negative ROS1 IHC results (hazard ratio [HR], 0.96; 95% confidence interval [CI], 0.54-1.72; P = .599) (Table 3). After factoring in age, tumor size, lymph node status, metastasis classification, lymphatic invasion, vascular invasion, and perineural invasion, positive ROS1 IHC status was not an independent prognostic factor (negative vs positive: HR, 1.16; 95% CI, 0.67-1.99; P = .811) (Table 3).
Table 3. Univariate and Multivariate Analyses of Prognostic Factors for Overall Survival
In the past several years, chromosomal rearrangements in RTKs also have been discovered in major epithelial tumors. such as NSCLC,5, 18 colon cancer,19, 20 and breast cancer.20 A recent report identified 2 fusion transcripts (neurogenic differentiation 2 and v-erb-b2 erythroblastic leukemia viral oncogene 2 [NEUROD2-ERRB2] and cyclin-dependent kinase 12 [CDK12]-ERBB2) in the MKN7 GC cell line.21 The CDK12-ERBB2 fusion transcript did not generate a full-length CDK12-ERBB2 fusion protein containing the HER2 kinase domain because of internal in-frame stop codons, which led only to the generation of a truncated CDK12 protein.21 By using an N-terminus monoclonal antibody, we identified 23 GC samples that exhibited strong (3+) perinuclear staining. These 23 ROS1 IHC GC samples exhibited lower lymph node status, poorly differentiated tumors, and a decreased tendency for lymphatic invasion. Moreover, all 23 samples were negative for HER2 amplification in FISH analysis. RT-PCR and break-apart FISH analyses of these 23 ROS1 IHC-positive GC samples revealed that 2 GC samples contained SLC34A2-ROS1 transcripts. These 2 patients presented with poorly differentiated, diffuse type adenocarcinoma in the antrum, and both died of the disease within 2 years from the date of surgery because of peritoneal recurrence. ROS1 protein is not normally expressed in the stomach8, 13; however, in a carcinogen-induced rat GC model, ROS1 was identified as 1 of the growth-promoting genes that remained overexpressed 4 weeks after cessation of the carcinogen exposure, indicating that aberrant ROS1 overexpression may play a role in the pathogenesis of GC.22
Currently, 8 fusion partners to ROS1(golgi-associated PDZ and coiled-coil motif containing [FIG], SLC34A2, CD74, SCD4, TPM3, LRIG3, EZR, and KDEL [Lys-Asp-Glu-Leu] endoplasmic reticulum protein retention receptor 2 [KDELR2]) have been identified in NSCLC.5-10 For glioblastoma multiforme and cholangiocarcinoma, only a FIG-ROS1 fusion transcript has been identified to date.4, 11 Thus, SLC34A2-ROS1 in GC represents the first reported ROS1 fusion variant in tumor other than NSCLC. SLC34A2 is an Na-dependent phosphate cotransporter and generally is not expressed in gastric tissues but is identified in lung, small intestine, and kidney tissues.23, 24 Thus, our discovery of SLC34A2-ROS1 in GC indicates that there may be baseline expression of SLC34A2 in the secretary epithelium of the stomach. SLC34A2-ROS1 is also the major fusion transcript identified in the HCC78 cell line.5 It has been demonstrated that SLC34A2-ROS1 is inhibited by ALK inhibitors (crizotinib6, 9, 25 and TAE68411) in vitro. Indeed, the use of crizotinib in ROS1 rearranged NSCLC has demonstrated significant clinical activity, opening up the possibility of using crizotinib and other ROS1 inhibitors in other tumors that harbor ROS1 rearrangement (National Clinical Trials [NCT] identification no. NCT00858195).12 Currently, in addition to crizotinib, 3 different ROS1 inhibitors (AP26311 [NCT01449461], ASP3062 [NCT1284192], and AZD1480 [NCT01 12397 and NCT01219543]) are in early phase clinical trials (available at: www.clinicaltrials.gov; accessed August 31, 2012). It remains to be determined whether SLC34A2-ROS1 is a driver mutation in GC. Not until there is systemic screening for ROS1 rearrangement in GC followed by the administration of ROS1 inhibitors in trials that demonstrate a clinical benefit for patients with ROS1-rearranged GC will there be proof of principle that ROS1 rearrangement is a driver mutation in GC, as has been the case for NSCLC.12, 26 Therefore, the discovery of an SLC34A2-ROS1 rearrangement in GC will spur further efforts to screen for ROS1 rearrangement in GC and to better characterize these patients.
The incidence of ROS1 rearrangement as detected by ROS1 break-apart FISH in GC in our study is estimated at 0.6% (3 of 495 samples). Our use of a commercially available N-terminal antibody may represent an underestimation of the incidence of ROS1-rearranged GC, because we may have missed the C-terminus of the fusion protein that contains the kinase domain. There is currently a proprietary C-terminal antibody, D4D6, which, when available commercially, eventually may be able to identify more ROS1-rearranged GC samples. Furthermore, the break-apart FISH criteria for detecting ROS1 rearrangement have not been standardized even in NSCLC. Bergethon et al.6 considered the separation of probes by greater than the width 2 signal diameters as positive, whereas Davies et al.8 considered a tumor sample ROS1 FISH-positive if the probes were separated 1 signal diameter. We used the more conservative criteria adapted for ALK break-apart FISH of a separation of probes greater than 2 signal diameters. Furthermore, the 15% cutoff value was adapted from the criteria for ALK rearrangement and has yet to be validated in ROS1-rearranged NSCLC or GC. We did identify a GC sample among all 23 ROS1 IHC-positive samples that contained 13% cells harboring break-apart ROS1 FISH, but no abnormal ROS1 fusion variants were identified.
Nevertheless, it is important to remember that the initial discovery of ROS1 rearrangement in NSCLC in 2007 was based on the identification in 1 cell line; and it is important to note that it was observed in only 1 patient sample.6 Despite our use of PCR primers for the other 5 known fusion partners to ROS1 in detecting other ROS1 fusions, we were not able to identify additional ROS1 fusions other than SLC34A2-ROS1. This may be because of false-positive ROS1 IHC tests, a lack of mRNA preservation in the extracted tissue, a low level of the expressed ROS1 fusion transcripts, or a combination of these factors. Indeed, among the 18 ROS1-rearranged NSCLCs described by Bergethon et al.,6 fusion partners were identified in only 6 tumors. It is not unreasonable to expect that all 8 fusions partners to ROS1 will be identified in GC.
It is noteworthy that both ROS1-rearranged GCs were negative for HER2 amplification by FISH and negative for MET overexpression by IHC, indicating that a ROS1 rearrangement may define a new molecular subtype of GC. However, because we identified only 3 patients who had ROS1-rearranged GC, it remains to define the full spectrum of clinicopathologic characteristics of these patients. In conclusion, for the first time, we identified a small subgroup of patients with GC who had ROS1 rearrangements confirmed by FISH and RT-PCR analyses. For patients with GC who have ROS1 rearrangements, this discovery may lead to a potential clinical benefit from ROS1 inhibitors through enriched molecular cohort clinical trials.
This study was supported by grants from the Korea Healthcare Technology R&D Project, Ministry for Health and Welfare Affairs, Republic of Korea (A092255 and A101130) and by Samsung Biomedical Research Institute grant SBRI-CB11031.
CONFLICT OF INTEREST DISCLOSURES
S.-H.I.O. received consulting honorarium and research funding from Pfizer.